Skip to content

Contributors & citation

Use this page to see who builds MVIS and how to cite it.

  • OpenMVIS: the project.
  • MVIS: the calibration toolbox, installed as the Python package pymvis.
  • Calibench: the datasets and results on Hugging Face.
  • Yulin Yang (yangyulin): author of MVIS, the regression suite and the benchmark; maintainer.
  • Yuantai Zhang (Yuantai-Z) and Xingxing Zuo (advisor): camera-only pipeline fixes (outlier rejection, pose alignment, frame selection), the ROS1 recalibration tool (tools/ros1_recalib) and RealSense D435i recordings for the Device wiki.
  • Nerako Lee (nerakoo) and Chuchu Chen (advisor): RealSense D455/D456 chessboard recordings and calibration runs for the Device wiki and the Calibench-VI community datasets.

The list follows AUTHORS in the MVIS repository. How we credit contributors is in Governance. The groups and companies who shared recordings are thanked on Acknowledgements.

One entry covers the toolbox, this site and the benchmark results. Its DOI is the Zenodo concept DOI, which resolves to the latest release; each release also has its own version DOI on Zenodo.

@misc{openmvis,
title = {{OpenMVIS}: open-source multi-visual-inertial sensor calibration},
author = {{OpenMVIS Contributors}},
year = {2026},
doi = {10.5281/zenodo.24239182},
url = {https://openmvis.com}
}

Results come in dated batches. If you compare against them, name the batch you used, so the comparison stays citable after later batches land:

OpenMVIS leaderboard, batch 2026-11-10a, accessed <date>. https://openmvis.com/benchmark/results/

The batch shown on the Leaderboard is the current one. Methodology explains what the numbers mean.

Cite the MVIS paper, the preferred citation in CITATION.cff:

@article{yang2024mvis,
title = {Multi-visual-inertial system: Analysis, calibration, and estimation},
author = {Yang, Yulin and Geneva, Patrick and Huang, Guoquan},
journal = {The International Journal of Robotics Research},
year = {2024},
volume = {43},
number = {13},
pages = {1995--2026},
doi = {10.1177/02783649241245726}
}

If you use analytic combined IMU integration, ACI2, or ACI3 (ACI2 with IMU intrinsics, from the MVIS paper above), also cite the ACI2 paper. Every MVIS camera–IMU solve uses ACI3 in its IMU factors.

@inproceedings{yang2020aci2,
title = {Analytic Combined {IMU} Integration ({ACI}$^2$) For Visual Inertial Navigation},
author = {Yang, Yulin and Wisely Babu, Benzun Pious and Chen, Chuchu and Huang, Guoquan and Ren, Liu},
booktitle = {2020 IEEE International Conference on Robotics and Automation (ICRA)},
year = {2020},
pages = {4680--4686},
doi = {10.1109/ICRA40945.2020.9197280}
}

When your results depend on a dataset, cite the dataset as well as MVIS.

  • VI-Rig recordings (our own): cite Calibench-VI for every VI-Rig set. For the 4-IMU sets (virig-4imus-3cams, virig-4imus-3cams-aruco and virig-4imus-4cams), also cite the MVIS paper above. virig-5imus-4cams has no paper yet; cite Calibench-VI only.
  • Third-party datasets: cite the original authors. The dataset card lists the citation each one requires.
@misc{calibench2026,
author = {Yang, Yulin},
title = {Calibench-VI: the visual-inertial calibration benchmark},
year = {2026},
howpublished = {Hugging Face Datasets},
note = {https://huggingface.co/datasets/calibench-vi/datasets}
}

The same entries are on Datasets.

MVIS implements the method of the MVIS paper (Yang, Geneva and Huang, IJRR 2024), written at the Robot Perception and Navigation Group (RPNG), University of Delaware.